Given the context, reference extraction, candidate extraction and evaluation result, generate the steps to evaluate whether the candidate extraction is valid or not.
Both the reference and the candidate represent the extraction of the causes, relation and effects from the context.
The extractions have the format `[Cause] <cause> [Relation] <relation> [Effect] <effect>`.
The candidate extraction is valid if has the same meaning as the reference.

Context:
{INPUT}

Reference extraction:
{GOLD}

Candidate extraction:
{OUTPUT}

{RESULT_MODE}: {RESULT}

Let's think step by step:
